1. Cochrane and the BMJ    The current edition of the British Medical Journal (1) discusses the joys of randomised, controlled clinical trials. The joy is relative, for these RCTs comprise a tiny minority of all published results of the investigations of medical treatments.

Randomised controlled trials may not be the last word in seeking after truth, but they do stand for best practicable levels of intellectual hygiene. They use methodologies more likely to distinguish between reality and appearance and which steer away from more and less conscious bias. RCTs seek to offer hard evidence of benefit and risk in place of clinical impressions, received wisdoms and wishful thinking. They are the better alternative to the "flood of deficiencies" that mark most formal investigations of clinical effectiveness:

"Trials are often too small, too short, of poor quality, and poorly resented, and they address the wrong question … Methodological inadequacies distort results … Few trials include adequate measures of quality of life … Cost data are poorly presented … The ethical aspects of trials are often neglected … The views of patients are either not sought or forgotten, and participants in trials often have limited understanding of what is happening … Trials are usually poorly managed … Politics have hijacked the conclusions of some trials … Marketeers can use trials to further their own profit making ends." (2)

The dedication of this issue of the BMJ notionally marks the 50^th anniversary of the publication of the 'first' randomised controlled trial - but implicitly it celebrates the recent accomplishments of the Cochrane Collaboration (3). Barely founded seven ago, the Collaboration has become a major international driving force for integrity and openness in clinical research. The strength of the movement (and to some extent its fragility) was underlined by the 600-strong attendance, from 34 countries, at the 6^th annual Cochrane Colloquium, held in Baltimore two weeks ago.

For love rather than money, a variety of Cochrane task forces are now engaged in the arduous process of sifting wheat from chaff - first identifying all investigations of treatment, then attempting systematic analyses of the reliable ones. The ratio of quality to quantity seems highly unsatisfactory: good standards of prescribing would be hard to sustain on such a low protein diet.

The Cochrane Collaboration on Depression, Anxiety and Neurosis has, for example, now compiled a database of 8,515 references of which 4,348 (51%) relate to depression. Of these, only 301 (7%) report the results of randomised controlled clinical trials. Most deal with treatment of major depression in hospital settings - though nine out of ten patients are treated by GPs. Seventy-three per cent of the trials evaluated drug therapy for no more than 6-7 weeks, and only about 4% lasted over 24-weeks - the usually recommended minimum period of treatment. (4, 5, 6)

The same sort of picture applies to other therapies - so much so that, in a recent address to the Royal College of Psychiatrists, the BMJ editor reportedly complained that most clinical trials were "rubbish" (7). On this occasion, however, Richard Smith marks the triumph of celebration over despair. His introduction concludes: "How exciting that there is so much room to improve".

2. A good move by Glaxo-Wellcome    Also in this issue of the BMJ is notice of an initiative by Glaxo-Wellcome that deserves a good round of applause (8). True, I have a rather a soft spot for them, but I applaud in spite of the company's commitment "to work in partnership with researchers, healthcare providers, and governments". Why not consumers too?

Nevertheless, it is good to see Glaxo-Wellcome taking a lead in disclosure, to the extent of now publishing on its website all protocols and results from clinical trials. This will indeed make it easier for independent researchers to undertake systematic reviews of trials. It will also help to improve standards of investigation, and communication about research being done.

When this new Glaxo-Wellcome system is up and running, there is another, obviously important and useful step the company could take. This would be to publish the names of all the investigators involved in developing protocols and running trials. On the face of it, this might seem a very sensitive issue, yet it is precisely this that makes the case for disclosure seem so strong.

The issue is sensitive because it touches on the extent of company influence in non-commercial sectors, and on possible conflicts of interest and commitment by independent researchers. But this is the very reason for disclosure - to give evidence there is nothing to hide. In the long run, disclosure would earn both investigators and the company an indispensable element of public trust. In the meantime, other companies would do well to follow this Glaxo Wellcome lead.

3. Much ado about Prozac   Four recently published trials, sponsored by Eli Lilly, all seek to make essentially the same point - one that the manufacturers have been drumming home for the last two years. The message is that, unlike major competing drugs - notably paroxetine and sertraline - fluoxetine has a "protective effect", sparing users bad symptoms on withdrawal.

Naturally, the question of dependence has not been discussed. To have done so would have involved [a] raising fundamental questions about the long-term effectiveness of antidepressants; [b] disregarding the conclusions of many trials, and the appropriateness of therapeutic guidelines based on them; [c] questioning the wisdom of the medical establishment, even raising the spectre of something akin to semantic fraud - the result of [d] decreeing that, however frequent, severe or persistent withdrawal symptoms might be, they no longer amount to "dependence" (as they did until a few years ago).

The first trial sets the scene for the other three. On the basis of this apparently well-controlled but clearly biased investigation, it was established that fluoxetine did not cause (acute and relatively visible) withdrawal symptoms of the kind often seen with competing brands. This finding permitted the remaining three trials to be conducted on the basis that "fluoxetine is not associated with withdrawal symptoms" - which in turn meant that, if a patient's mood was seen to deteriorate after stopping fluoxetine, it should always be counted as a sign of relapse into depression, and therefore as evidence the drug really worked.

This claim seems dubious, since the study involved placebo substitution for active drug only for 5-8 days. This period of interruption of treatment would show up withdrawal effects with sertraline and paroxetine (having shorter half-lives), much more than with fluoxetine (which is much more slowly cleared from the body). Nevertheless, the study is based on very interesting data - perhaps the strongest evidence yet of the magnitude of a dependence risk, at least with other SSRIs.

2. J Zajecka, M D Fawcett, J Amsterdam, F Quitkin, F Reimherr, J Rosenbaum, D Michelson, C Beasley: Safety of Abrupt Discontinuation of Fluoxetine: A Randomised, Placebo-Controlled Study, J. Clin Psychopharmacol., 1998, 18, 3, 193-197.